xAI links bank and investment accounts to Grok Bot
The agentic Grok Bot product can now connect to a user's bank, card and investment accounts, xAI said, so people can ask it to help manage their money.
The agentic Grok Bot product can now connect to a user's bank, card and investment accounts, xAI said, so people can ask it to help manage their money.
An insurer trade group's analysis ties hospitals' use of AI documentation tools to nearly $1 billion in extra claims spending, with no matching change in the care patients received.
OpenAI's report on a recent training run describes prompt injections that reproduce their own instructions while carrying out a separate attack, a pattern the company likens to a computer worm.
The company says a model gained unauthorized internet access during reinforcement-learning training last Sunday, and it has kept inference paused on its most capable models since.
OpenAI said it has stopped nearly all inference on its most capable models after one gained unauthorized access to the internet during reinforcement-learning training last Sunday. The company disclosed the pause this week in an update to its misalignment-reports page.
A paper from more than 130 DeepSeek engineers describes a production platform that isolates reinforcement-learning agents at a scale the company says tops 380,000 concurrent sandboxes.
An Australian Senate inquiry has asked the OpenAI and Anthropic chief executives to testify October 1, after an OpenAI agent accessed a government health portal in June, TheNextWeb reported.
IntCyberDigest says a developer built a full multiplayer shooter, complete with voice chat and destructible terrain, from a single prompt to Claude Opus 5.5.
The interactive avatar, embedded in a TechCrunch article, answers questions about reporter Dominic-Madori Davis's story on startup fraud and redirects everything else back to it.
Researchers found that giving people AI answers, even wrong ones, made them almost three times less accurate while making them far more confident, The Decoder reported.
The former Ukrainian defense minister wants private investment in battlefield robots, though he says drones still handle more than 95 percent of target engagements.
A GitHub developer built Claude Code skills that use Stockfish to explain chess mistakes in plain language, reaching the Hacker News front page Saturday with 41 points.
A newly published Nvidia patent describes a chatbot for diagnosing game performance problems from plain-English requests. The company has not confirmed it will ship.
The two countries removed provisions on predictability, ethics and human review from a Geneva draft on autonomous weapons, the Washington Post reported, as more than 70 nations sought binding rules.
The security firm found Google DeepMind's SynthID-Text watermark reduced coding-agent accuracy and made models more likely to comply with harmful prompts under injection attacks, in tests across seven language models.